ECMWF model accuracy


If you download the ECMWF Hires model and archive it on a daily basis as I do from, you can then create a grid of forecast charts from ten separate runs (T+240 down to T+24) for a specific date and time. In this example a grid of ten forecast frames for 00 UTC on the 1st of April, which I top of with a final frame of the analysis from the UKMO to compare them all against. ‘What the hell for?’ I hear you shout. Well I think it’s a very easy way to verify how accurate the world’s leading NWP model is, and show it’s limitations using the mark one human eyeball. Years ago I did some verification of various NWP models in which I concluded that you couldn’t trust any of them at T+120 or beyond. I suppose it all depends on the weather regime, but it maybe that the best models can be trusted to give a decent forecast these days at T+120 or maybe even at T+144. That seems to be the case with this series of runs for the 1st of April from the ECMWF.

1 thought on “ECMWF model accuracy”

  1. I keep meaning to do something similar, but only ever get as far as remembering what the weather was supposed to be doing on a given day. So, thanks!

What do you think?

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Scroll to Top
%d bloggers like this: